{"id":"https://openalex.org/W7080313836","doi":"https://doi.org/10.1109/tvlsi.2025.3604419","title":"DiffAccel: Accelerating Diffusion Models Through Adaptive Feature Optimization and Dynamic Hardware Adaptation","display_name":"DiffAccel: Accelerating Diffusion Models Through Adaptive Feature Optimization and Dynamic Hardware Adaptation","publication_year":2025,"publication_date":"2025-09-08","ids":{"openalex":"https://openalex.org/W7080313836","doi":"https://doi.org/10.1109/tvlsi.2025.3604419"},"language":"en","primary_location":{"id":"doi:10.1109/tvlsi.2025.3604419","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tvlsi.2025.3604419","pdf_url":null,"source":{"id":"https://openalex.org/S37538908","display_name":"IEEE Transactions on Very Large Scale Integration (VLSI) Systems","issn_l":"1063-8210","issn":["1063-8210","1557-9999"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Very Large Scale Integration (VLSI) Systems","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Enhao Tang","orcid":"https://orcid.org/0000-0002-0539-8885"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Enhao Tang","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China"],"raw_orcid":"https://orcid.org/0000-0002-0539-8885","affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Weize Ma","orcid":"https://orcid.org/0009-0008-2365-5284"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Weize Ma","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China"],"raw_orcid":"https://orcid.org/0009-0008-2365-5284","affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Yudan Jiang","orcid":null},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yudan Jiang","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Sheng Xu","orcid":null},"institutions":[{"id":"https://openalex.org/I76569877","display_name":"Southeast University","ror":"https://ror.org/04ct4d772","country_code":"CN","type":"education","lineage":["https://openalex.org/I76569877"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Sheng Xu","raw_affiliation_strings":["School of Instrument Science and Engineering, Southeast University, Nanjing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Instrument Science and Engineering, Southeast University, Nanjing, China","institution_ids":["https://openalex.org/I76569877"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Zhongfeng Wang","orcid":"https://orcid.org/0000-0002-7227-4786"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhongfeng Wang","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China"],"raw_orcid":"https://orcid.org/0000-0002-7227-4786","affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"last","author":{"id":null,"display_name":"Jun Lin","orcid":"https://orcid.org/0009-0005-3505-4847"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jun Lin","raw_affiliation_strings":["School of Electronic Science and Engineering and the Interdisciplinary Research Center for Future Intelligent Chips (Chip-X), Nanjing University, Nanjing, China"],"raw_orcid":"https://orcid.org/0009-0005-3505-4847","affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering and the Interdisciplinary Research Center for Future Intelligent Chips (Chip-X), Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.46276566,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"33","issue":"12","first_page":"3290","last_page":"3303"},"is_retracted":false,"is_paratext":false,"is_xpac":true,"primary_topic":{"id":"https://openalex.org/T12157","display_name":"Geochemistry and Geologic Mapping","score":0.6453999876976013,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T12157","display_name":"Geochemistry and Geologic Mapping","score":0.6453999876976013,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T13067","display_name":"Geological Modeling and Analysis","score":0.024700000882148743,"subfield":{"id":"https://openalex.org/subfields/1906","display_name":"Geochemistry and Petrology"},"field":{"id":"https://openalex.org/fields/19","display_name":"Earth and Planetary Sciences"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T14311","display_name":"Electrical and Electromagnetic Research","score":0.024000000208616257,"subfield":{"id":"https://openalex.org/subfields/3107","display_name":"Atomic and Molecular Physics, and Optics"},"field":{"id":"https://openalex.org/fields/31","display_name":"Physics and Astronomy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.5849000215530396},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.5741999745368958},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5627999901771545},{"id":"https://openalex.org/keywords/dataflow","display_name":"Dataflow","score":0.5300999879837036},{"id":"https://openalex.org/keywords/flexibility","display_name":"Flexibility (engineering)","score":0.4936000108718872},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.4724999964237213},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.3822000026702881},{"id":"https://openalex.org/keywords/adaptation","display_name":"Adaptation (eye)","score":0.3815000057220459},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.3521000146865845}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8039000034332275},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.5849000215530396},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.5741999745368958},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5627999901771545},{"id":"https://openalex.org/C96324660","wikidata":"https://www.wikidata.org/wiki/Q205446","display_name":"Dataflow","level":2,"score":0.5300999879837036},{"id":"https://openalex.org/C2780598303","wikidata":"https://www.wikidata.org/wiki/Q65921492","display_name":"Flexibility (engineering)","level":2,"score":0.4936000108718872},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.49079999327659607},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.4724999964237213},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.4074000120162964},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.3822000026702881},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.3815000057220459},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.37439998984336853},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.3521000146865845},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.34439998865127563},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.3262999951839447},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.3246000111103058},{"id":"https://openalex.org/C13164978","wikidata":"https://www.wikidata.org/wiki/Q600158","display_name":"Hardware acceleration","level":3,"score":0.314300000667572},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.305400013923645},{"id":"https://openalex.org/C80469333","wikidata":"https://www.wikidata.org/wiki/Q189088","display_name":"Von Neumann architecture","level":2,"score":0.2971999943256378},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2964000105857849},{"id":"https://openalex.org/C2779304628","wikidata":"https://www.wikidata.org/wiki/Q3503480","display_name":"Face (sociological concept)","level":2,"score":0.28780001401901245},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.28110000491142273},{"id":"https://openalex.org/C2781357197","wikidata":"https://www.wikidata.org/wiki/Q5757597","display_name":"High memory","level":2,"score":0.2800000011920929},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.27869999408721924},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.2784999907016754},{"id":"https://openalex.org/C118702147","wikidata":"https://www.wikidata.org/wiki/Q189396","display_name":"Dynamic random-access memory","level":3,"score":0.27070000767707825},{"id":"https://openalex.org/C20136886","wikidata":"https://www.wikidata.org/wiki/Q749647","display_name":"Interoperability","level":2,"score":0.2705000042915344},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.26919999718666077},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.2685999870300293},{"id":"https://openalex.org/C127162648","wikidata":"https://www.wikidata.org/wiki/Q16858953","display_name":"Channel (broadcasting)","level":2,"score":0.2563999891281128},{"id":"https://openalex.org/C176649486","wikidata":"https://www.wikidata.org/wiki/Q2308807","display_name":"Memory management","level":3,"score":0.2526000142097473}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tvlsi.2025.3604419","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tvlsi.2025.3604419","pdf_url":null,"source":{"id":"https://openalex.org/S37538908","display_name":"IEEE Transactions on Very Large Scale Integration (VLSI) Systems","issn_l":"1063-8210","issn":["1063-8210","1557-9999"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Very Large Scale Integration (VLSI) Systems","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2250357272","display_name":null,"funder_award_id":"62174084","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5354445887","display_name":null,"funder_award_id":"62341408","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":33,"referenced_works":["https://openalex.org/W1861492603","https://openalex.org/W1901129140","https://openalex.org/W2194775991","https://openalex.org/W2913292019","https://openalex.org/W2962834855","https://openalex.org/W3034600949","https://openalex.org/W3034771037","https://openalex.org/W3035204081","https://openalex.org/W3153469116","https://openalex.org/W3184059350","https://openalex.org/W3216352822","https://openalex.org/W4226014430","https://openalex.org/W4250482878","https://openalex.org/W4312933868","https://openalex.org/W4385245566","https://openalex.org/W4386057725","https://openalex.org/W4386071831","https://openalex.org/W4386072096","https://openalex.org/W4387449305","https://openalex.org/W4389610328","https://openalex.org/W4390874074","https://openalex.org/W4391305822","https://openalex.org/W4392746069","https://openalex.org/W4401211860","https://openalex.org/W4402475889","https://openalex.org/W4402715897","https://openalex.org/W4402778510","https://openalex.org/W4403278953","https://openalex.org/W4404134143","https://openalex.org/W4408151358","https://openalex.org/W4408182767","https://openalex.org/W4408183360","https://openalex.org/W4409248685"],"related_works":[],"abstract_inverted_index":{"Diffusion":[0],"models":[1],"(DMs)":[2],"have":[3],"emerged":[4],"as":[5],"a":[6,167,202,309],"revolutionary":[7],"technology":[8],"in":[9,32],"AI-generated":[10],"content.":[11],"Despite":[12],"their":[13,33,46],"superior":[14],"quality,":[15],"DMs":[16],"face":[17,64],"significant":[18],"deployment":[19],"challenges":[20],"due":[21],"to":[22,106,108,130,173,218,248,294,298,347],"two":[23,138],"key":[24,139,244],"limitations:":[25],"1)":[26],"the":[27,39,43,73,96,102,109,116,132,164,194,221,227,275,299],"substantial":[28],"computational":[29,53,111,196,290],"cost":[30],"inherent":[31],"iterative":[34],"denoising":[35],"process":[36],"and":[37,55,69,87,104,112,151,188,214,234,253,265,288,341],"2)":[38],"complexities":[40],"of":[41,45,134,198,226,314],"optimizing":[42],"execution":[44,176],"multiscale":[47],"U-Net":[48,179,245],"architecture,":[49],"which":[50],"exhibits":[51],"diverse":[52,178],"patterns":[54,197],"varying":[56],"memory":[57,113,204,224,232],"requirements":[58],"across":[59,177],"layers.":[60],"Existing":[61],"optimization":[62],"approaches":[63],"limitations":[65],"on":[66,158,193,318],"both":[67],"algorithmic":[68,140],"hardware":[70,97,165,315],"fronts.":[71],"At":[72,274],"algorithm":[74,276],"level,":[75,166,277],"current":[76,99],"methods":[77],"struggle":[78],"with":[79,91,308],"inefficient":[80],"feature":[81,89,145,154],"enhancement":[82,146],"through":[83],"uniform":[84],"channel":[85],"processing":[86,259],"inflexible":[88],"reuse":[90],"static":[92],"caching":[93,155],"strategies.":[94],"On":[95],"side,":[98],"solutions":[100],"lack":[101],"flexibility":[103],"efficiency":[105,233,345],"adapt":[107],"varied":[110],"demands":[114],"within":[115],"U-Net.":[117],"To":[118],"address":[119],"these":[120],"challenges,":[121],"we":[122],"propose":[123],"DiffAccel,":[124],"an":[125,152],"algorithm\u2013hardware":[126],"co-optimized":[127],"accelerator":[128],"designed":[129],"accelerate":[131],"inference":[133],"various":[135],"DMs.":[136],"First,":[137],"innovations":[141],"are":[142],"introduced:":[143],"selective":[144],"targeting":[147],"only":[148],"high-dimensional":[149],"channels":[150],"adaptive":[153,211],"mechanism":[156],"based":[157,192,317],"identified":[159],"evolution":[160],"patterns.":[161],"Second,":[162],"at":[163],"reconfigurable":[168],"array":[169],"(RA)":[170],"is":[171,208,241,260,268,306],"proposed":[172],"enable":[174],"efficient":[175],"layers":[180],"by":[181],"switching":[182],"its":[183],"dataflow":[184],"between":[185],"output-stationary":[186],"(OS)":[187],"weight-stationary":[189],"(WS)":[190],"modes":[191],"specific":[195],"each":[199],"layer.":[200],"Moreover,":[201],"data-volume-aware":[203],"management":[205],"unit":[206],"(MMU)":[207],"implemented,":[209],"employing":[210],"input":[212],"allocation":[213],"hierarchical":[215],"output":[216,267],"storage":[217],"efficiently":[219],"manage":[220],"disparate":[222],"layer":[223],"footprints":[225],"U-Net,":[228],"thereby":[229],"improving":[230],"on-chip":[231],"reducing":[235],"waste.":[236],"Third,":[237],"fine-grained":[238],"pipeline":[239],"fusion":[240],"employed":[242],"for":[243,262,270],"operation":[246],"sequences":[247],"minimize":[249],"intermediate":[250],"data":[251],"movement":[252],"significantly":[254],"reduce":[255],"latency.":[256],"Specifically,":[257],"row-first":[258],"applied":[261],"linear\u2013Softmax":[263],"operations,":[264],"channel-priority":[266],"utilized":[269],"convolution\u2013groupnorm":[271],"(GNorm)":[272],"sequences.":[273],"DiffAccel":[278,305,324],"achieves":[279,325],"<inline-formula":[280,326],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[281,327,333],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">":[282,328,334],"<tex-math":[283,329,335],"notation=\"LaTeX\">$2.93\\times":[284],"$</tex-math>":[285,331,337],"</inline-formula>":[286,338],"speedup":[287],"reduces":[289],"complexity":[291],"(from":[292],"105":[293],"69":[295],"TFLOPs)":[296],"compared":[297,346],"baseline":[300],"stable":[301],"diffusion":[302],"(SD)":[303],"v2.0.":[304],"implemented":[307],"28-nm":[310],"technology.":[311],"In":[312],"terms":[313],"efficiency,":[316],"post":[319],"place-and-route":[320],"(P&R)":[321],"implementation":[322],"results,":[323],"notation=\"LaTeX\">$5.25\\times":[330],"</inline-formula>\u2013<inline-formula":[332],"notation=\"LaTeX\">$5.79\\times":[336],"higher":[339],"throughput":[340],"maintains":[342],"comparable":[343],"area":[344],"state-of-the-art":[348],"(SOTA)":[349],"accelerators.":[350]},"counts_by_year":[],"updated_date":"2025-11-27T01:12:40.094763","created_date":"2025-10-10T00:00:00"}